Moments

On March 16 this year, I picked up my WeChat public account again to publish articles and images. The last time I posted was on March 28, 2020, and it had been abandoned for six years.

If I had spent those six years cultivating a public account, that persistent writing would itself have been a harvest.

Six years ago, I had no awareness or understanding of the compounding effect of time. Over these six years, my family has gained a lot—we've traveled many places and seen many sights. But I haven't felt much personal growth or accumulation. Many times, I just thought about things in my mind, took in a lot of input, but never produced any output.

Learning, practicing, and producing output form a good closed-loop learning path. I believe this loop should not rely on self-discipline or persistence, but rather on recognizing it as a scientific method worth adopting.

Now, I write down the thoughts in my mind. In the past, I used to think it was enough to just know something—writing it out seemed too troublesome, and what if I didn't write well? In short, various worries in my heart caused delays, and then nothing happened.

Recently, as I've been writing more, my habit has started to change: just write directly without overthinking, because words are so easy to correct and revise.

When I'm typing words, I quickly enter a calm state of mind, as if I'm having a conversation with myself, and time seems to disappear. Writing does take up some of my free time allowance, but isn't output itself also a form of freedom (freedom of expression)?

“Validate fast, fail fast” — I finally feel like I truly understand this mindset now, which is essentially the MVP approach.

My old process used to be: confirm the idea, choose the tech stack, build the architecture, design the UI, and dive deep into details. Most of my time and energy were consumed by the technical side upfront, rather than validating the idea itself or testing feasibility. In the end, many of those projects still failed. It wasn’t a complete waste, but overall it wasn’t an efficient way to build products.

Recently, I’ve started shifting my mindset and practicing the MVP approach instead. I focus first on validating the core functionality. If the product works and the direction makes sense, then I can refine and polish it later. This saves a huge amount of time and energy, while also giving me feedback as quickly as possible.

The time you truly control, and the lasting sense of ease and freedom you can sustain, are the best rewards you can give yourself.

Codex mobile is here — and it’s blowing up!

I’d recommend a code agent product I’ve been using recently: OpenClacky. I topped up $20 and have been using the DeepSeek V4 Flash model. So far, it’s been surprisingly cheap — I haven’t even used up $1 a day.

One issue I’ve run into though: when the agent is actively working in the terminal, if I try to enter a new task at the same time, the whole thing just freezes and stops responding for quite a while. So now I usually wait for the current task to finish before sending the next instruction.

Huashu Design , Say one sentence to your agent — Claude Code, Cursor, Codex, OpenClaw, Hermes all work.

3 to 30 minutes — you ship a product launch animation, a clickable App prototype, an editable PPT deck, a print-grade infographic.

Not "decent for AI" quality — it looks like a real design team made it. Give the skill your brand assets (logo, colors, UI screenshots) and it reads your brand's voice; give it nothing and the built-in 20 design vocabularies still keep you out of AI slop territory.

teco-lab.com, this website for learning spoken English through English vlogs is really well done. A lifetime membership only costs RMB 39.9. Since I watch English vlogs quite often, I had actually thought about building a similar English learning website myself. But now it seems I don’t need to anymore.

heroui.com - Beautiful, fast and modern React UI library. (Previously NextUI)

fumadocs.dev is a beautiful and flexible documentation framework for any React.js project.

A few days ago, I tried translating a PDF ebook using Ollama with a local Qwen 3.5:4b model. Translation speed wasn't an issue, but the final layout turned out poorly. My workflow was to first convert the PDF to Markdown, translate it, and then generate a new PDF. However, the bilingual vertical layout (Chinese-English) in the resulting PDF was hard to read and very uncomfortable.

I looked into this problem and realized PDF text layout is indeed difficult to handle. Different books have different content and require different presentation styles. Optimizing and adjusting the layout for each book costs too much effort.

Finally, I found that HTML is the most suitable choice—styling is easy and results in a clean layout, and my goal is simply to translate and read comfortably. With this approach, I'll continue to improve it later and use it to help me read English books.